Diagnosis in Automotive Applications A Case Study with the Model Compilation Approach

This paper addresses intricate diagnosis situations in modern cars. It employs a rather new and little known diagnosis paradigm, which is a mixture of model-based and heuristic diagnosis and which is called model compilation [5, 25, 26]. The basic idea is to simulate a model of the system under study in various fault modes and over its typical input range to compile a simulation database C. A simplified rule-based behavior model is constructed from C, in which long cause-effect chains are replaced with much simpler associations and which is optimized for heuristic classification of the fault being studied. The main contribution of the paper is the application of this idea to the complex discrete event / continuous system models of modern cars. Our experiments, which are based on the latest simulation models from the automotive industry, are encouraging and may indicate a paradigm shift away from the diagnosis approaches used so far in this field.